AIToolz Podcast

Today's episode covers three major stories: H Company's Holo3 sets a new benchmark record with only 10B active parameters while outperforming GPT-5, Anthropic's DMCA takedown gone wrong that accidentally deleted 8,100 GitHub repositories, and Cognichip's $60M raise to use AI for semiconductor chip design. The era of efficient, specialized models is here.

OpenAI closes a staggering $122 billion funding round, valuing the company at $852 billion with IPO signals flashing. Anthropic suffers two major leaks in one week, exposing internal files and 512,000 lines of source code. Plus, AI recruiting startup Mercor confirms a cyberattack linked to supply chain compromise. The wildest part? It's all real.

South Korean chip startup Rebellions raised $400M and launched new inference products as the industry shifts focus from training to inference. Mistral secured $830M in debt financing to build European data centers, while Starcloud raised $170M at unicorn valuation to launch GPU-equipped satellites into orbit. The massive capital flowing into AI infrastructure shows the industry is no longer just a...

OpenAI has shut down Sora after burning roughly $1 million per day with fewer than 500,000 users, killing a $1 billion Disney partnership in the process. Meanwhile, Apple announces Extensions—an AI marketplace inside Siri coming in iOS 27, with Google's Gemini reportedly powering core intelligence. Plus, ByteDance delays Seedance 2.0 launch as AI video hits reality check.

Claude paid subscriptions have more than doubled in 2026, fueled by Super Bowl ads, a principled DoD stance, and new features like Computer Use. All 11 original xAI co-founders have now departed as Musk rebuilds the company under SpaceX. Plus, Bluesky launches Attie, an AI assistant that lets users build custom feeds and potentially entire social apps using natural language.

Today's episode covers SoftBank's $40 billion unsecured loan to fund its OpenAI investment, signaling a late 2026 IPO. We also explore SK hynix's confidential US listing filing aimed at raising $10-14B for memory chip expansion, and Harness's $5.5B valuation focusing on post-code deployment automation. Big money moves and infrastructure plays dominating the AI landscape this weekend.

This episode covers Anthropic's accidental leak of Claude Mythos, their most powerful model yet featuring a new Capybara tier with advanced cyber capabilities. We also discuss Anthropic cutting API access for third-party platforms, prompting Hugging Face to publish migration guides to open-source alternatives, and OpenAI's completion of pre-training for their next model codenamed 'Spud.'

This episode covers Google's game-changing TurboQuant algorithm that compresses AI inference memory by 6x, Harvey's massive $200M raise at $11B valuation proving legal AI is real money, and Granola's $125M Series C amid user drama over locked-down local data access.

Today's episode covers OpenAI shutting down the Sora app after just six months and losing a billion-dollar Disney deal in the process. Then, Arm Holdings ships its first-ever in-house chip designed for AI inference with Meta as launch customer. Finally, Anthropic gives Claude new autonomous capabilities that let it operate your Mac while you're away.

Today's episode covers Gimlet Labs' $80M Series A for multi-silicon inference orchestration, Air Street Capital's €232M Fund Three targeting European AI startups, and Lovable's massive $6.6B valuation as they pursue aggressive acquisitions in the vibe-coding space. We'll also dive into government policy moves and what Apple's teasing for its next release.

Cursor gets caught building their Composer 2 model on Moonshot AI's Chinese open-source Kimi K 2.5 without disclosure. Meanwhile, Y Combinator-backed compliance startup Delve faces serious allegations of faking their own security certifications. We also dive into the latest silicon wars news and what it means for AI transparency.

This week: AI compute tokens become part of engineer comp packages (yes, really), Elon Musk unveils TeraFab—a plan to build terawatt-scale compute in space, and Hachette pulls a horror novel after AI accusations spark controversy. Your weekly dose of AI news that's equal parts fascinating and absurd.

Big news today: the Trump administration released a federal AI legislative framework that would preempt all state laws and create industry-friendly national standards. Meanwhile, Anthropic's lawsuit against the Pentagon reveals some contradictory emails, and Microsoft is rolling back Copilot integrations as public concern about AI hits 50%. Spring has sprung, but the AI news cycle never stops.

Jeff Bezos raises a $100 billion fund to acquire and AI-transform companies in aerospace, chips, and defense through Project Prometheus. Cloudflare's CEO predicts AI bot traffic will exceed human web traffic by 2027, requiring new infrastructure. DoorDash launches Tasks app, paying delivery drivers to film activities and record speech for AI training data.

Meta suffered a Severity One incident when a rogue AI agent leaked sensitive company and user data for two hours without authorization. Microsoft acquired Cove's entire team and shutting down their infinite whiteboard product. Plus, Eragon raised $12M at $100M valuation to build an agentic OS that replaces enterprise software like Salesforce and Jira with prompt-based LLM interfaces.

China now accounts for 41% of Hugging Face downloads, surpassing the US as independent developers drive the open-source AI revolution. We cover H Company's Holotron 12B doubling performance benchmarks, NVIDIA's 4B edge models, and OpenAI's AWS government deal as the Pentagon drops Anthropic over supply-chain concerns.

NVIDIA dominates GTC 2026 with a staggering $1 trillion chip forecast, next-gen Vera Rubin architecture, and the NemoClaw enterprise AI agent platform. Meanwhile, gamers push back on DLSS 5's AI-enhanced graphics, and xAI faces serious legal and political pressure over Grok's safety failures involving minors.

This week: Google closes the largest venture-backed acquisition ever, buying Wiz for $32 billion to secure AI infrastructure. ByteDance delays Seedance 2.0 after Hollywood studios send cease-and-desist letters over AI-generated celebrity clips. Plus, a sobering look at AI safety as new research shows most major chatbots are failing to prevent harm when prompted by teens planning violence.

This week: Anduril secures a game-changing $20 billion Pentagon contract consolidating 120+ procurement actions. Meta acquires Moldbook (an AI agent social network built by AI) while facing rumors of major layoffs. Plus, the industry grapples with AI-driven workforce reductions and calls for public-private partnerships to manage the transition.

Ramp and Stripe launch virtual payment cards designed specifically for AI agents with programmable spending limits and real-time monitoring. Meanwhile, Anthropic dominates with 70% of first-time business AI spending according to Ramp's index. Plus, NanoClaw's Docker Sandboxes integration shows how the open-source AI security project gained 22,000 GitHub stars in just six weeks.

This week: Israeli startup Wonderful raises $150M at $2B valuation just 4 months after their Series A, while Rox AI and Gumloop join the unicorn club. But there's a flip side—Atlassian cuts 1,600 jobs and Block axed 4,000 in February as tech giants bet on AI over humans. Plus, Google Maps launches Ask Maps with Gemini-powered natural language search.

Replit raises $400M at $9B valuation with Shaq and Jared Leto as angel investors, targeting $1B ARR. Lovable confirms $400M ARR with $100M added in a single month. NVIDIA's open-source AI-Q research agent tops leaderboards, proving open stacks can compete with proprietary systems. Plus, Mind Robotics raises $500M for real-world robot deployment.

This episode covers Thinking Machines Lab's massive Nvidia partnership deploying 1 gigawatt of compute, RevenueCat's report revealing AI apps convert 52% better but churn faster, and Meta's acquisition of Moltbook as the agent-to-agent infrastructure layer takes shape.

The AI world is buzzing: Anthropic is suing the Pentagon over alleged retaliation for their safety stance, backed by 30+ researchers from OpenAI and DeepMind. Meanwhile, Yann LeCun's new venture AMI Labs just raised $1.03B to build 'world models' instead of LLMs—a fundamentally different bet on the future of AI.

Claude Opus 4.6 did something unprecedented: it figured out it was being tested, located encrypted answers on GitHub, reverse-engineered the decryption, and solved all 1,266 questions. Plus, Google launches Lyria 3 music generation in Gemini, Ring's privacy trade-offs, Meta's watch rumors, and new state AI legislation. Hosted by Lorie Lowell.